Summary Background Comprehensive long-term data on atrial fibrillation trends in men and women are scant. We aimed to provide such data through analysis of the Framingham cohort over 50 years. Methods We investigated trends in incidence, prevalence, and risk factors for atrial fibrillation and its association with stroke and mortality after onset in 9511 participants enrolled in the Framingham Heart Study between 1958 and 2007. We analysed trends within 10 year groups (1958–67, 1968–77, 1978–87, 1988–97, and 1998–2007), stratified by sex. Findings During 50 years of observation (202 417 person-years), 1544 cases of new-onset atrial fibrillation occurred (of whom 723 47% were women). Between 1958–67 and 1998–2007, age-adjusted prevalence of atrial fibrillation quadrupled from 20·4 to 96·2 cases per 1000 person-years in men and from 13·7 to 49·4 cases per 1000 person-years in women; age-adjusted incidence increased from 3·7 to 13·4 new cases per 1000 person-years in men and from 2·5 to 8·6 new cases per 1000 person-years in women (ptrend <0·0001 for all comparisons). For atrial fibrillation diagnosed by electrocardiograph (ECG) during routine Framingham examinations, age-adjusted prevalence per 1000 person-years increased (12·6 in 1958–67 to 25·7 in 1998–2007 in men, ptrend =0·0007; 8·1 to 11·8 in women, ptrend =0·009). However, age-adjusted incidence of atrial fibrillation by Framingham Heart Study ECGs did not change significantly with time. Although the prevalence of most risk factors changed over time, their associated hazards for atrial fibrillation changed little. Multivariable-adjusted proportional hazards models revealed a 74% (95% CI 50–86%) decrease in stroke (hazards ratio HR 3·77, 95% CI 1·98–7·20 in 1958–1967 compared with 1998–2007; ptrend =0·0001) and a 25% (95% CI −3–46%) decrease in mortality (HR 1·34, 95% CI 0·97–1·86 in 1958–1967 compared with 1998–2007; ptrend =0·003) in 20 years following atrial fibrillation onset. Interpretation Trends of increased incidence and prevalence of atrial fibrillation in the community were probably partly due to enhanced surveillance. Measures are needed to enhance early detection of atrial fibrillation, through increased awareness coupled with targeted screening programmes and risk factor-specific prevention. Funding NIH, NHLBI, NINDS, Deutsche Forschungsgemeinschaft.